22M:132

Fall 07
J. Simon
Comments on
Quotient Spaces and Quotient Maps
There are many situations in topology where we build a topological space by
starting with some (often simpler) space[s] and doing some kind of gluing or
identications. The situations may look dierent at rst, but really they are
instances of the same general construction. In the rst section below, we give some
examples, without any explanation of the theoretical/technial issues. In the next
section, we give the general denition of a quotient space and examples of several
kinds of constructions that are all special instances of this general one.
1. Examples of building topological spaces with interesting shapes
by starting with simpler spaces and doing some kind of gluing or
identifications.
Example 0.1. Identify the two endpoints of a line segment to form a circle.
Example 0.2. Identify two opposite edges of a rectangle (i.e. a rectangular strip of
paper) to form a cylinder.
c J. Simon, all rights reserved page 1
Example 0.3. Now identify the top and bottom circles of the cylinder to each other,
resulting in a 2-dimensional surface called a torus.
Example 0.4. Start with a round 2-dimensional disk; identify the whole boundary
circle to a single point. The result is a surface, the 2-sphere.
Example 0.5. ?? Take two dierent spheres (in particular, disjoint from each
other); pick one point on each sphere and glue the two spheres together by
identifying the chosen point from each.
c J. Simon, all rights reserved page 2
Example 0.6. The set of all homeomorphisms from a space X onto itself forms a
group: the operation is composition (The composition of two homeomorphisms is a
homeomorphism; functional composition is associative; the identity map I : X X
is the identity element in the group of homeomorphisms.) The group of all
self-homeomorphisms of X may have interesting subgroups. When we specify some
[sub]group of homeomorphisms of X that is isomorphic to some abstract group G,
we call this an action of the group G on X. Note that in this situation, we are
viewing the elements of G as homeomorphisms of X, and the group operation in G
as function composition; so, in particular, for each g, h G, x X, we are insisting
that (g h)(x) = g(h(x)).
When we have a group G acting on a space X, there is a natural quotient space.
For each x X, let Gx = {g(x) | g G}. View each of these orbit sets as a single
point in some new space X

.
2. Definition of quotient space
Suppose X is a topological space, and suppose we have some equivalence relation
dened on X. Let X

be the set of equivalence classes. We want to dene a

special topology on X, called the quotient topology. To do this, it is convenient to
introduce the function
: X X

dened by
(x) = [x] ,
that is, (x) = the equivalence class containing x.This can be confusing, so say it
over to yourself a few times: is a function from X into the power set of X; it
assigns to each point x X a certain subset of X, namely the equivalence class
containing the point x. Since each x X is contained in exactly one equivalence
class, the function x [x] is well-dened. At the risk of belaboring the obvious,
since each equivalence class has at least one member, the function is surjective.
Before talking about the quotient topology, lets look at several examples of the
quotient sets X

.
Example 0.7. Let X = {a, b, c}, a set with 3 points. Partition X into two
equivalence classes: {1, 3} and {2}. So X

has two elements, call them O and T.

The function is
(1) = O, (2) = T, (3) = O .
Example 0.8. In our previous example 0.1, one equivalence class has two elements;
every other equivalence class is a singleton. Likewise, in example ??, one
equivalence class has two elements and all the others are singletons.
Example 0.9. In our previous example 0.2, each equivalence class coming from
points on the vertical edges being identied consists of two points; all other
equivalence classes are singletons.
c J. Simon, all rights reserved page 3
Example 0.10 (shrinking a set to a point). Let X be any space, and A X. The
the quotient space X/A is the set of equivalence classes [x], where [x] = A if x A
and [x] = {x} if x / A. The set X

has one giant point A and the rest are just

the points of X A. This is the situation in example 0.4
Example 0.11. Let X be the real line R
1
. Let G be the additive group of integers,
Z. Dene an action of G on R by n(x) = x + n for each n Z, x R. Here the set
Gx consists of all integer translates of the point x. Note that the sets Gx do form a
partition of X. That is, the relation x y there exists g G with g(x) = y is
an equivalence relation on X. [Unassigned exercise: Check this claim; it depends on
the fact that G is a group.]
However, unlike our previous examples, it may be not so obvious what a geometric
picture of X

looks like: the number of points is the same as the half-open interval
[0, 1); but what should the topology be??
3. The quotient topology
If we think of constructing X

by actually picking up a set X and squishing some

parts together, we would like the passage from X to X

to be continuous. We make
this precise by insisting that the projection map
: X X

(x) = [x]
be continuous. This puts an obligation on the topology we assign to X

: If a set U
is open in X

then
1
(U) is open in X. (Think of this as an upper bound on
which sets in X

can be open.) We dene the quotient topology on X

by letting all
sets U that pass this test be admitted.
A set U is open in X

if and only if
1
(U) is open in X. The quotient topology on
X

is the nest topology on X

for which the projection map is continuous.

We now have an unambiguously dened special topology on the set X

of
equivalence classes. But that does not mean that it is easy to recognize which
topology is the right one. Going back to our example 0.6, the set of equivalence
classes (i.e. orbit sets Gx) is in 1-1 correspondence with the points of the half-open
interval [0, 1). But that does not imply that the quotient space, with the quotient
topology, is homeomorphic to the usual [0, 1). To understand how to recognize the
quotient spaces, we introduce the idea of quotient map and then develop the texts
Theorem 22.2. This theorem may look cryptic, but it is the tool we use to prove
that when we think we know what a quotient space looks like, we are right (or to
help discover that our intuitive answer is wrong).
c J. Simon, all rights reserved page 4
4. Quotient maps
Suppose p : X Y is a map such that
a . p is surjective,
b . p is continuous [i.e. U open in Y = p
1
open in X], and
c . U Y , p
1
(U) open in X = U open in Y .
In this case we say the map p is a quotient map.
The last two items say that U is open in Y if and only if p
1
(U) is open in X.
Theorem. If p : X Y is surjective, continuous, and an open map, the p is a
quotient map. If p : X Y is surjective, continuous, and a closed map, then p is a
quotient map.
Proof. The proof of this theorem is left as an unassigned exercise; it is not hard, and
you should know how to do it. (Consider this part of the list of sample problems for
the next exam.)
Remark. Note that the properties open map and closed map are independent of
each other (there are maps that are one but not the other) and strictly stronger than
quotient map [HW Exercise 3 page 145].
Remark (Saturated sets). A quotient map does not have to be an open map. But it
does have the property that certain open sets in X are taken to open sets in Y . We
say that a set V X is saturated with respect to a function f [or with respect to an
equivalence relation ] if V is a union of point-inverses [resp. union of equivalence
classes]. A quotient map has the property that the image of a saturated open set is
open. Likewise, when dening the quotient topology, the function : X X

takes
saturated open sets to open sets.
Suppose now that you have a space X and an equivalence relation . You form the
set of equivalence classes X

and you give X

the quotient topology. How can you

know what the space X

looks like? Lets use the group action example 0.6 to

illustrate how we can answer the question.
The interval [0, 1) has the right number of points. So does the circle S
1
. Which is
really the quotient space? It turns out S
1
is the right answer. To see this, dene a
map q : R S
1
by q(t) = cos(2t), sin(2t) R
2
. We claim this function q has
two properties; together, they imply (Theorem 22.2) that the quotient space of R
under the action of Z is homeomorphic to S
1
.
a . The function q distinguishes points of the domain R exactly the same way
as the equivalence relation . That is, q(x) = q(y) [x] = [y].
b . The map q is a quotient map.
Proposition. In the preceding example (action of Z on R), X

is homeomorphic to
S
1
.
c J. Simon, all rights reserved page 5
Proof. Dene a function f : S
1
X

as follows: For each z S

1
, f(z) = (q
1
(z)).
Because of condition (a) above, the function f is well-dened and is a bijection
between S
1
and X

.
We next show f is continuous. Let U be an open set in X

. Then
1
(U) is an open
set in R that is saturated with respect to . By condition (a), since and q have
exactly the same point-inverses,
1
(U) is also saturated with respect to q. But
then, since q is a quotient map, q(
1
(U)) is open in S
1
. Since f
1
(U) is precisely
q(
1
(U)), we have that f
1
(U) is open. The proof that f
1
is continuous is almost
identical.

There is one case of quotient map that is particularly easy to recognize. Once we
study compact spaces, we will have the following:
Theorem. Suppose f : X Y is a continuous surjective function. If X is compact
and Y is Hausdor, then f is a quotient map.
Proof. We show that f is a closed map. Let C be a closed subset of X. A closed
subset of a compact space is compact, so C is compact. The continuous image of a
compact set is compact, so f(C) is a compact subset of Y . A compact subset of a
Hausdor space is closed. So f(C) is closed in Y .
This theorem tells us that in all of our examples above (except 0.6, where we needed
a fancier proof), we have the right picture of the quotient space.
[end of handout]
c J. Simon, all rights reserved page 6